From: Neil Horman <nhorman@redhat.com>
Date: Thu, 12 Jul 2007 14:47:07 -0400
Subject: [net] sctp: rewrite receive buffer management code
Message-id: 20070712184706.GF12979@hmsendeavour.rdu.redhat.com
O-Subject: [RHEL 5.2 PATCH] rewrite of sctp receive buffer management code (bz 246722)
Bugzilla: 246722
Hey all-
For a long time now, we've had a pretty serious problem with sctp, in
that its easy to make it start dropping frames due to receive buffer limitations
before its receive window closes. This leads to all sorts of bad behavior that
we've had to hack around for quite some time, both in RHEL and upstream. A few
weeks ago, I wrote this patch to replace all those hack with reasonable receive
and send buffer management that is in line with the way tcp does memory
management:
http://git.kernel.org/?p=linux/kernel/git/vxy/lksctp-2.6.23.git;a=commit;h=178e3a7bf7892b4e7589d99329c3318866ea27c0
Its been tested throughly by me and the sctp maintainer upstream, and is slated
for inclusion to 2.6.23. This is a backport of that patch, and its supporting
bits. It resolves bz 246722, and should likely solve several other sctp
performance related bz's.
Thanks & Regards
Neil
Acked-by: "David S. Miller" <davem@redhat.com>
---
include/linux/sysctl.h | 3 +
include/net/sctp/sctp.h | 19 +++++++
include/net/sctp/ulpevent.h | 1 +
net/sctp/endpointola.c | 1 +
net/sctp/protocol.c | 32 +++++++++++
net/sctp/sm_statefuns.c | 73 +++++++--------------------
net/sctp/socket.c | 120 +++++++++++++++++++++++++++++++++++++-----
net/sctp/sysctl.c | 33 ++++++++++++
net/sctp/ulpevent.c | 36 ++++++++++---
net/sctp/ulpqueue.c | 3 +-
10 files changed, 242 insertions(+), 79 deletions(-)
diff --git a/include/linux/sysctl.h b/include/linux/sysctl.h
index 9bf1545..909fd90 100644
--- a/include/linux/sysctl.h
+++ b/include/linux/sysctl.h
@@ -775,6 +775,9 @@ enum {
NET_SCTP_SNDBUF_POLICY = 15,
NET_SCTP_SACK_TIMEOUT = 16,
NET_SCTP_RCVBUF_POLICY = 17,
+ NET_SCTP_BUF_MEM = 18,
+ NET_SCTP_BUF_RMEM = 19,
+ NET_SCTP_BUF_WMEM = 20,
};
/* /proc/sys/net/bridge */
diff --git a/include/net/sctp/sctp.h b/include/net/sctp/sctp.h
index 92eae0e..178300f 100644
--- a/include/net/sctp/sctp.h
+++ b/include/net/sctp/sctp.h
@@ -137,6 +137,7 @@ int sctp_inet_listen(struct socket *sock, int backlog);
void sctp_write_space(struct sock *sk);
unsigned int sctp_poll(struct file *file, struct socket *sock,
poll_table *wait);
+void sctp_sock_rfree(struct sk_buff *skb);
/*
* sctp/primitive.c
@@ -398,6 +399,24 @@ static inline struct list_head *sctp_list_dequeue(struct list_head *list)
return result;
}
+/* SCTP version of skb_set_owner_r. We need this one because
+ * of the way we have to do receive buffer accounting on bundled
+ * chunks.
+ */
+static inline void sctp_skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
+{
+ struct sctp_ulpevent *event = sctp_skb2event(skb);
+
+ skb->sk = sk;
+ skb->destructor = sctp_sock_rfree;
+ atomic_add(event->rmem_len, &sk->sk_rmem_alloc);
+ /*
+ * This mimics the behavior of
+ * sk_stream_set_owner_r
+ */
+ sk->sk_forward_alloc -= event->rmem_len;
+}
+
/* Tests if the list has one and only one entry. */
static inline int sctp_list_single_entry(struct list_head *head)
{
diff --git a/include/net/sctp/ulpevent.h b/include/net/sctp/ulpevent.h
index 6c40cfc..1a4ddc1 100644
--- a/include/net/sctp/ulpevent.h
+++ b/include/net/sctp/ulpevent.h
@@ -63,6 +63,7 @@ struct sctp_ulpevent {
__u32 cumtsn;
int msg_flags;
int iif;
+ unsigned int rmem_len;
};
/* Retrieve the skb this event sits inside of. */
diff --git a/net/sctp/endpointola.c b/net/sctp/endpointola.c
index ffda1d6..cd85b79 100644
--- a/net/sctp/endpointola.c
+++ b/net/sctp/endpointola.c
@@ -101,6 +101,7 @@ static struct sctp_endpoint *sctp_endpoint_init(struct sctp_endpoint *ep,
/* Use SCTP specific send buffer space queues. */
ep->sndbuf_policy = sctp_sndbuf_policy;
+
sk->sk_write_space = sctp_write_space;
sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
diff --git a/net/sctp/protocol.c b/net/sctp/protocol.c
index 1ab03a2..8762957 100644
--- a/net/sctp/protocol.c
+++ b/net/sctp/protocol.c
@@ -51,6 +51,7 @@
#include <linux/netdevice.h>
#include <linux/inetdevice.h>
#include <linux/seq_file.h>
+#include <linux/bootmem.h>
#include <net/protocol.h>
#include <net/ip.h>
#include <net/ipv6.h>
@@ -89,6 +90,10 @@ extern int sctp_eps_proc_exit(void);
extern int sctp_assocs_proc_init(void);
extern int sctp_assocs_proc_exit(void);
+extern int sysctl_sctp_mem[3];
+extern int sysctl_sctp_rmem[3];
+extern int sysctl_sctp_wmem[3];
+
/* Return the address of the control sock. */
struct sock *sctp_get_ctl_sock(void)
{
@@ -982,6 +987,8 @@ SCTP_STATIC __init int sctp_init(void)
int i;
int status = -EINVAL;
unsigned long goal;
+ unsigned long limit;
+ int max_share;
int order;
/* SCTP_DEBUG sanity check. */
@@ -1086,6 +1093,31 @@ SCTP_STATIC __init int sctp_init(void)
/* Initialize handle used for association ids. */
idr_init(&sctp_assocs_id);
+ /* Set the pressure threshold to be a fraction of global memory that
+ * is up to 1/2 at 256 MB, decreasing toward zero with the amount of
+ * memory, with a floor of 128 pages.
+ * Note this initalizes the data in sctpv6_prot too
+ * Unabashedly stolen from tcp_init
+ */
+ limit = min(num_physpages, 1UL<<(28-PAGE_SHIFT)) >> (20-PAGE_SHIFT);
+ limit = (limit * (num_physpages >> (20-PAGE_SHIFT))) >> (PAGE_SHIFT-11);
+ limit = max(limit, 128UL);
+ sysctl_sctp_mem[0] = limit / 4 * 3;
+ sysctl_sctp_mem[1] = limit;
+ sysctl_sctp_mem[2] = sysctl_sctp_mem[0] * 2;
+
+ /* Set per-socket limits to no more than 1/128 the pressure threshold*/
+ limit = (sysctl_sctp_mem[1]) << (PAGE_SHIFT - 7);
+ max_share = min(4UL*1024*1024, limit);
+
+ sysctl_sctp_rmem[0] = PAGE_SIZE; /* give each asoc 1 page min */
+ sysctl_sctp_rmem[1] = (1500 *(sizeof(struct sk_buff) + 1));
+ sysctl_sctp_rmem[2] = max(sysctl_sctp_rmem[1], max_share);
+
+ sysctl_sctp_wmem[0] = SK_STREAM_MEM_QUANTUM;
+ sysctl_sctp_wmem[1] = 16*1024;
+ sysctl_sctp_wmem[2] = max(64*1024, max_share);
+
/* Size and allocate the association hash table.
* The methodology is similar to that of the tcp hash tables.
*/
diff --git a/net/sctp/sm_statefuns.c b/net/sctp/sm_statefuns.c
index 5b5ae79..a4a2037 100644
--- a/net/sctp/sm_statefuns.c
+++ b/net/sctp/sm_statefuns.c
@@ -5174,7 +5174,6 @@ static int sctp_eat_data(const struct sctp_association *asoc,
int account_value;
struct sctp_tsnmap *map = (struct sctp_tsnmap *)&asoc->peer.tsn_map;
struct sock *sk = asoc->base.sk;
- int rcvbuf_over = 0;
data_hdr = chunk->subh.data_hdr = (sctp_datahdr_t *)chunk->skb->data;
skb_pull(chunk->skb, sizeof(sctp_datahdr_t));
@@ -5184,48 +5183,6 @@ static int sctp_eat_data(const struct sctp_association *asoc,
/* ASSERT: Now skb->data is really the user data. */
- /*
- * If we are established, and we have used up our receive buffer
- * memory, think about droping the frame.
- * Note that we have an opportunity to improve performance here.
- * If we accept one chunk from an skbuff, we have to keep all the
- * memory of that skbuff around until the chunk is read into user
- * space. Therefore, once we accept 1 chunk we may as well accept all
- * remaining chunks in the skbuff. The data_accepted flag helps us do
- * that.
- */
- if ((asoc->state == SCTP_STATE_ESTABLISHED) && (!chunk->data_accepted)) {
- /*
- * If the receive buffer policy is 1, then each
- * association can allocate up to sk_rcvbuf bytes
- * otherwise, all the associations in aggregate
- * may allocate up to sk_rcvbuf bytes
- */
- if (asoc->ep->rcvbuf_policy)
- account_value = atomic_read(&asoc->rmem_alloc);
- else
- account_value = atomic_read(&sk->sk_rmem_alloc);
- if (account_value > sk->sk_rcvbuf) {
- /*
- * We need to make forward progress, even when we are
- * under memory pressure, so we always allow the
- * next tsn after the ctsn ack point to be accepted.
- * This lets us avoid deadlocks in which we have to
- * drop frames that would otherwise let us drain the
- * receive queue.
- */
- if ((sctp_tsnmap_get_ctsn(map) + 1) != tsn)
- return SCTP_IERROR_IGNORE_TSN;
-
- /*
- * We're going to accept the frame but we should renege
- * to make space for it. This will send us down that
- * path later in this function.
- */
- rcvbuf_over = 1;
- }
- }
-
/* Process ECN based congestion.
*
* Since the chunk structure is reused for all chunks within
@@ -5285,18 +5242,9 @@ static int sctp_eat_data(const struct sctp_association *asoc,
* seems a bit troublesome in that frag_point varies based on
* PMTU. In cases, such as loopback, this might be a rather
* large spill over.
- * NOTE: If we have a full receive buffer here, we only renege if
- * our receiver can still make progress without the tsn being
- * received. We do this because in the event that the associations
- * receive queue is empty we are filling a leading gap, and since
- * reneging moves the gap to the end of the tsn stream, we are likely
- * to stall again very shortly. Avoiding the renege when we fill a
- * leading gap is a good heuristic for avoiding such steady state
- * stalls.
- */
- if (!asoc->rwnd || asoc->rwnd_over ||
- (datalen > asoc->rwnd + asoc->frag_point) ||
- (rcvbuf_over && (!skb_queue_len(&sk->sk_receive_queue)))) {
+ */
+ if ((!chunk->data_accepted) && (!asoc->rwnd || asoc->rwnd_over ||
+ (datalen > asoc->rwnd + asoc->frag_point))) {
/* If this is the next TSN, consider reneging to make
* room. Note: Playing nice with a confused sender. A
@@ -5317,6 +5265,21 @@ static int sctp_eat_data(const struct sctp_association *asoc,
}
/*
+ * Also try to renege to limit our memory usage in the event that
+ * we are under memory pressure
+ * If we can't renege, don't worry about it, the sk_stream_rmem_schedule
+ * in sctp_ulpevent_make_rcvmsg will drop the frame if we grow our
+ * memory usage too much
+ */
+ if (*sk->sk_prot_creator->memory_pressure) {
+ if (sctp_tsnmap_has_gap(map) &&
+ (sctp_tsnmap_get_ctsn(map) + 1) == tsn) {
+ SCTP_DEBUG_PRINTK("Under Pressure! Reneging for tsn:%u\n", tsn);
+ deliver = SCTP_CMD_RENEGE;
+ }
+ }
+
+ /*
* Section 3.3.10.9 No User Data (9)
*
* Cause of error
diff --git a/net/sctp/socket.c b/net/sctp/socket.c
index dab1594..3309778 100644
--- a/net/sctp/socket.c
+++ b/net/sctp/socket.c
@@ -109,23 +109,42 @@ static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
extern kmem_cache_t *sctp_bucket_cachep;
+extern int sysctl_sctp_mem[3];
+extern int sysctl_sctp_rmem[3];
+extern int sysctl_sctp_wmem[3];
+
+int sctp_memory_pressure;
+atomic_t sctp_memory_allocated;
+atomic_t sctp_sockets_allocated;
+
+static void sctp_enter_memory_pressure(void)
+{
+ sctp_memory_pressure = 1;
+}
+
/* Get the sndbuf space available at the time on the association. */
static inline int sctp_wspace(struct sctp_association *asoc)
{
+ int amt;
struct sock *sk = asoc->base.sk;
- int amt = 0;
- if (asoc->ep->sndbuf_policy) {
- /* make sure that no association uses more than sk_sndbuf */
- amt = sk->sk_sndbuf - asoc->sndbuf_used;
+ if (asoc->ep->sndbuf_policy)
+ amt = asoc->sndbuf_used;
+ else
+ amt = atomic_read(&sk->sk_wmem_alloc);
+
+ if (amt >= sk->sk_sndbuf) {
+ if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
+ amt = 0;
+ else {
+ amt = sk_stream_wspace(sk);
+ if (amt < 0)
+ amt = 0;
+ }
} else {
- /* do socket level accounting */
- amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
+ amt = sk->sk_sndbuf - amt;
}
-
- if (amt < 0)
- amt = 0;
-
+
return amt;
}
@@ -157,6 +176,7 @@ static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
sizeof(struct sctp_chunk);
atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
+ sk_charge_skb(sk, chunk->skb);
}
/* Verify that this is a valid address. */
@@ -3121,6 +3141,7 @@ SCTP_STATIC int sctp_init_sock(struct sock *sk)
sp->hmac = NULL;
SCTP_DBG_OBJCNT_INC(sock);
+ atomic_inc(&sctp_sockets_allocated);
return 0;
}
@@ -3134,7 +3155,7 @@ SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
/* Release our hold on the endpoint. */
ep = sctp_sk(sk)->ep;
sctp_endpoint_free(ep);
-
+ atomic_dec(&sctp_sockets_allocated);
return 0;
}
@@ -5357,12 +5378,38 @@ static void sctp_wfree(struct sk_buff *skb)
atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
+ /*
+ * This undoes what is done via sk_charge_skb
+ */
+ sk->sk_wmem_queued -= skb->truesize;
+ sk->sk_forward_alloc += skb->truesize;
+
sock_wfree(skb);
__sctp_write_space(asoc);
sctp_association_put(asoc);
}
+/* Do accounting for the receive space on the socket.
+ * Accounting for the association is done in ulpevent.c
+ * We set this as a destructor for the cloned data skbs so that
+ * accounting is done at the correct time.
+ */
+void sctp_sock_rfree(struct sk_buff *skb)
+{
+ struct sock *sk = skb->sk;
+ struct sctp_ulpevent *event = sctp_skb2event(skb);
+
+ atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
+
+ /*
+ * Mimic the behavior of sk_stream_rfree
+ */
+ sk->sk_forward_alloc += event->rmem_len;
+
+}
+
+
/* Helper function to wait for space in the sndbuf. */
static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
size_t msg_len)
@@ -5581,6 +5628,36 @@ void sctp_wait_for_close(struct sock *sk, long timeout)
finish_wait(sk->sk_sleep, &wait);
}
+static void sctp_sock_rfree_frag(struct sk_buff *skb)
+{
+ struct sk_buff *frag;
+
+ if (!skb->data_len)
+ goto done;
+
+ /* Don't forget the fragments. */
+ for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
+ sctp_sock_rfree_frag(frag);
+
+done:
+ sctp_sock_rfree(skb);
+}
+
+static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
+{
+ struct sk_buff *frag;
+
+ if (!skb->data_len)
+ goto done;
+
+ /* Don't forget the fragments. */
+ for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
+ sctp_skb_set_owner_r_frag(frag, sk);
+
+done:
+ sctp_skb_set_owner_r(skb, sk);
+}
+
/* Populate the fields of the newsk from the oldsk and migrate the assoc
* and its messages to the newsk.
*/
@@ -5633,10 +5710,10 @@ static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
event = sctp_skb2event(skb);
if (event->asoc == assoc) {
- sock_rfree(skb);
+ sctp_sock_rfree_frag(skb);
__skb_unlink(skb, &oldsk->sk_receive_queue);
__skb_queue_tail(&newsk->sk_receive_queue, skb);
- skb_set_owner_r(skb, newsk);
+ sctp_skb_set_owner_r_frag(skb, newsk);
}
}
@@ -5664,10 +5741,10 @@ static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
event = sctp_skb2event(skb);
if (event->asoc == assoc) {
- sock_rfree(skb);
+ sctp_sock_rfree_frag(skb);
__skb_unlink(skb, &oldsp->pd_lobby);
__skb_queue_tail(queue, skb);
- skb_set_owner_r(skb, newsk);
+ sctp_skb_set_owner_r_frag(skb, newsk);
}
}
@@ -5704,6 +5781,7 @@ static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
sctp_release_sock(newsk);
}
+
/* This proto struct describes the ULP interface for SCTP. */
struct proto sctp_prot = {
.name = "SCTP",
@@ -5726,6 +5804,12 @@ struct proto sctp_prot = {
.unhash = sctp_unhash,
.get_port = sctp_get_port,
.obj_size = sizeof(struct sctp_sock),
+ .sysctl_mem = sysctl_sctp_mem,
+ .sysctl_rmem = sysctl_sctp_rmem,
+ .sysctl_wmem = sysctl_sctp_wmem,
+ .memory_pressure = &sctp_memory_pressure,
+ .enter_memory_pressure = sctp_enter_memory_pressure,
+ .memory_allocated = &sctp_memory_allocated,
};
#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
@@ -5750,5 +5834,11 @@ struct proto sctpv6_prot = {
.unhash = sctp_unhash,
.get_port = sctp_get_port,
.obj_size = sizeof(struct sctp6_sock),
+ .sysctl_mem = sysctl_sctp_mem,
+ .sysctl_rmem = sysctl_sctp_rmem,
+ .sysctl_wmem = sysctl_sctp_wmem,
+ .memory_pressure = &sctp_memory_pressure,
+ .enter_memory_pressure = sctp_enter_memory_pressure,
+ .memory_allocated = &sctp_memory_allocated,
};
#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */
diff --git a/net/sctp/sysctl.c b/net/sctp/sysctl.c
index dc6f3ff..a97570e 100644
--- a/net/sctp/sysctl.c
+++ b/net/sctp/sysctl.c
@@ -51,6 +51,15 @@ static long rto_timer_max = 86400000; /* One day */
static long sack_timer_min = 1;
static long sack_timer_max = 500;
+int sysctl_sctp_mem[3];
+int sysctl_sctp_rmem[3];
+int sysctl_sctp_wmem[3];
+
+/*
+ * per assoc memory limitationf for sends
+ */
+int sysctl_sctp_wmem[3];
+
static ctl_table sctp_table[] = {
{
.ctl_name = NET_SCTP_RTO_INITIAL,
@@ -206,6 +215,30 @@ static ctl_table sctp_table[] = {
.extra1 = &sack_timer_min,
.extra2 = &sack_timer_max,
},
+ {
+ .ctl_name = NET_SCTP_BUF_MEM,
+ .procname = "sctp_mem",
+ .data = &sysctl_sctp_mem,
+ .maxlen = sizeof(sysctl_sctp_mem),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec,
+ },
+ {
+ .ctl_name = NET_SCTP_BUF_RMEM,
+ .procname = "sctp_rmem",
+ .data = &sysctl_sctp_rmem,
+ .maxlen = sizeof(sysctl_sctp_rmem),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec,
+ },
+ {
+ .ctl_name = NET_SCTP_BUF_WMEM,
+ .procname = "sctp_wmem",
+ .data = &sysctl_sctp_wmem,
+ .maxlen = sizeof(sysctl_sctp_wmem),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec,
+ },
{ .ctl_name = 0 }
};
diff --git a/net/sctp/ulpevent.c b/net/sctp/ulpevent.c
index ee23678..e7c7080 100644
--- a/net/sctp/ulpevent.c
+++ b/net/sctp/ulpevent.c
@@ -55,10 +55,12 @@ static void sctp_ulpevent_release_frag_data(struct sctp_ulpevent *event);
/* Initialize an ULP event from an given skb. */
-SCTP_STATIC void sctp_ulpevent_init(struct sctp_ulpevent *event, int msg_flags)
+SCTP_STATIC void sctp_ulpevent_init(struct sctp_ulpevent *event, int msg_flags,
+ unsigned int len)
{
memset(event, 0, sizeof(struct sctp_ulpevent));
event->msg_flags = msg_flags;
+ event->rmem_len = len;
}
/* Create a new sctp_ulpevent. */
@@ -73,7 +75,7 @@ SCTP_STATIC struct sctp_ulpevent *sctp_ulpevent_new(int size, int msg_flags,
goto fail;
event = sctp_skb2event(skb);
- sctp_ulpevent_init(event, msg_flags);
+ sctp_ulpevent_init(event, msg_flags, skb->truesize);
return event;
@@ -101,8 +103,8 @@ static inline void sctp_ulpevent_set_owner(struct sctp_ulpevent *event,
sctp_association_hold((struct sctp_association *)asoc);
skb = sctp_event2skb(event);
event->asoc = (struct sctp_association *)asoc;
- atomic_add(skb->truesize, &event->asoc->rmem_alloc);
- skb_set_owner_r(skb, asoc->base.sk);
+ atomic_add(event->rmem_len, &event->asoc->rmem_alloc);
+ sctp_skb_set_owner_r(skb, asoc->base.sk);
}
/* A simple destructor to give up the reference to the association. */
@@ -111,7 +113,7 @@ static inline void sctp_ulpevent_release_owner(struct sctp_ulpevent *event)
struct sctp_association *asoc = event->asoc;
struct sk_buff *skb = sctp_event2skb(event);
- atomic_sub(skb->truesize, &asoc->rmem_alloc);
+ atomic_sub(event->rmem_len, &asoc->rmem_alloc);
sctp_association_put(asoc);
}
@@ -372,7 +374,7 @@ struct sctp_ulpevent *sctp_ulpevent_make_remote_error(
/* Embed the event fields inside the cloned skb. */
event = sctp_skb2event(skb);
- sctp_ulpevent_init(event, MSG_NOTIFICATION);
+ sctp_ulpevent_init(event, MSG_NOTIFICATION, skb->truesize);
sre = (struct sctp_remote_error *)
skb_push(skb, sizeof(struct sctp_remote_error));
@@ -464,7 +466,7 @@ struct sctp_ulpevent *sctp_ulpevent_make_send_failed(
/* Embed the event fields inside the cloned skb. */
event = sctp_skb2event(skb);
- sctp_ulpevent_init(event, MSG_NOTIFICATION);
+ sctp_ulpevent_init(event, MSG_NOTIFICATION, skb->truesize);
ssf = (struct sctp_send_failed *)
skb_push(skb, sizeof(struct sctp_send_failed));
@@ -655,6 +657,24 @@ struct sctp_ulpevent *sctp_ulpevent_make_rcvmsg(struct sctp_association *asoc,
struct sctp_ulpevent *event = NULL;
struct sk_buff *skb;
size_t padding, len;
+ int rx_count;
+
+ /*
+ * check to see if we need to make space for this
+ * new skb, expand the rcvbuffer if needed, or drop
+ * the frame
+ */
+ if (asoc->ep->rcvbuf_policy)
+ rx_count = atomic_read(&asoc->rmem_alloc);
+ else
+ rx_count = atomic_read(&asoc->base.sk->sk_rmem_alloc);
+
+ if (rx_count >= asoc->base.sk->sk_rcvbuf) {
+
+ if ((asoc->base.sk->sk_userlocks & SOCK_RCVBUF_LOCK) ||
+ (!sk_stream_rmem_schedule(asoc->base.sk, chunk->skb)))
+ goto fail;
+ }
/* Clone the original skb, sharing the data. */
skb = skb_clone(chunk->skb, gfp);
@@ -683,7 +703,7 @@ struct sctp_ulpevent *sctp_ulpevent_make_rcvmsg(struct sctp_association *asoc,
event = sctp_skb2event(skb);
/* Initialize event with flags 0. */
- sctp_ulpevent_init(event, 0);
+ sctp_ulpevent_init(event, 0, skb->len + sizeof(struct sk_buff));
sctp_ulpevent_receive_data(event, asoc);
diff --git a/net/sctp/ulpqueue.c b/net/sctp/ulpqueue.c
index 575e556..8857ade 100644
--- a/net/sctp/ulpqueue.c
+++ b/net/sctp/ulpqueue.c
@@ -309,7 +309,7 @@ static struct sctp_ulpevent *sctp_make_reassembled_event(struct sk_buff_head *qu
if (!new)
return NULL; /* try again later */
- new->sk = f_frag->sk;
+ sctp_skb_set_owner_r(new, f_frag->sk);
skb_shinfo(new)->frag_list = pos;
} else
@@ -870,6 +870,7 @@ void sctp_ulpq_renege(struct sctp_ulpq *ulpq, struct sctp_chunk *chunk,
sctp_ulpq_partial_delivery(ulpq, chunk, gfp);
}
+ sk_stream_mem_reclaim(asoc->base.sk);
return;
}
--
1.5.3.5.645.gbb47
distrib
>
Scientific%20Linux
>
5x
>
x86_64
>
by-pkgid
>
27922b4260f65d317aabda37e42bbbff
>
files
>
3088
